This is the mail archive of the gcc-regression@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

A recent patch increased GCC's memory consumption!


Hi,

I am a friendly script caring about memory consumption in GCC.  Please
contact jh@suse.cz if something is going wrong.

Comparing memory consumption on compilation of combine.i, insn-attrtab.i,
and generate-3.4.ii I got:


comparing empty function compilation at -O0 level:
    Overall memory needed: 8152k -> 8153k
    Peak memory use before GGC: 1103k
    Peak memory use after GGC: 1000k
    Maximum of released memory in single GGC run: 130k
    Garbage: 239k
    Leak: 957k
    Overhead: 80k
    GGC runs: 2
Testing has produced no results
Testing has produced no results

comparing empty function compilation at -O0 -g level:
    Overall memory needed: 8168k -> 8169k
    Peak memory use before GGC: 1130k
    Peak memory use after GGC: 1035k
    Maximum of released memory in single GGC run: 129k
    Garbage: 241k
    Leak: 989k
    Overhead: 85k
    GGC runs: 4
Testing has produced no results
Testing has produced no results

comparing empty function compilation at -O1 level:
    Overall memory needed: 8204k -> 8205k
    Peak memory use before GGC: 1103k
    Peak memory use after GGC: 1000k
    Maximum of released memory in single GGC run: 132k
    Garbage: 241k
    Leak: 957k
    Overhead: 81k
    GGC runs: 2
    Pre-IPA-Garbage: 228k
    Pre-IPA-Leak: 960k
    Pre-IPA-Overhead: 79k
    Post-IPA-Garbage: 228k
    Post-IPA-Leak: 960k
    Post-IPA-Overhead: 79k

comparing empty function compilation at -O2 level:
    Overall memory needed: 8224k -> 8225k
    Peak memory use before GGC: 1103k
    Peak memory use after GGC: 1000k
    Maximum of released memory in single GGC run: 135k
    Garbage: 245k
    Leak: 958k
    Overhead: 82k
    GGC runs: 2
    Pre-IPA-Garbage: 228k
    Pre-IPA-Leak: 960k
    Pre-IPA-Overhead: 79k
    Post-IPA-Garbage: 228k
    Post-IPA-Leak: 960k
    Post-IPA-Overhead: 79k

comparing empty function compilation at -O3 level:
    Overall memory needed: 8224k -> 8225k
    Peak memory use before GGC: 1103k
    Peak memory use after GGC: 1000k
    Maximum of released memory in single GGC run: 135k
    Garbage: 245k
    Leak: 958k
    Overhead: 82k
    GGC runs: 2
    Pre-IPA-Garbage: 228k
    Pre-IPA-Leak: 960k
    Pre-IPA-Overhead: 79k
    Post-IPA-Garbage: 228k
    Post-IPA-Leak: 960k
    Post-IPA-Overhead: 79k

comparing combine.c compilation at -O0 level:
  Amount of produced GGC garbage increased from 37848k to 37894k, overall 0.12%
    Overall memory needed: 22640k -> 22641k
    Peak memory use before GGC: 7808k -> 7810k
    Peak memory use after GGC: 7189k
    Maximum of released memory in single GGC run: 1550k
    Garbage: 37848k -> 37894k
    Leak: 5517k
    Overhead: 4635k -> 4637k
    GGC runs: 372
Testing has produced no results
Testing has produced no results

comparing combine.c compilation at -O0 -g level:
    Overall memory needed: 24668k -> 24673k
    Peak memory use before GGC: 9648k
    Peak memory use after GGC: 8948k
    Maximum of released memory in single GGC run: 1838k
    Garbage: 38203k -> 38215k
    Leak: 8383k
    Overhead: 5263k -> 5264k
    GGC runs: 343
Testing has produced no results
Testing has produced no results

comparing combine.c compilation at -O1 level:
  Amount of memory still referenced at the end of compilation increased from 5656k to 5663k, overall 0.13%
    Overall memory needed: 32840k -> 32853k
    Peak memory use before GGC: 16263k -> 16268k
    Peak memory use after GGC: 16080k -> 16084k
    Maximum of released memory in single GGC run: 1379k -> 1380k
    Garbage: 51569k -> 51611k
    Leak: 5656k -> 5663k
    Overhead: 5805k -> 5810k
    GGC runs: 439 -> 438
    Pre-IPA-Garbage: 13925k -> 13937k
    Pre-IPA-Leak: 17092k -> 17097k
    Pre-IPA-Overhead: 2119k -> 2120k
    Post-IPA-Garbage: 13925k -> 13937k
    Post-IPA-Leak: 17092k -> 17097k
    Post-IPA-Overhead: 2119k -> 2120k

comparing combine.c compilation at -O2 level:
    Overall memory needed: 34660k -> 34661k
    Peak memory use before GGC: 16363k -> 16368k
    Peak memory use after GGC: 16197k -> 16201k
    Maximum of released memory in single GGC run: 1315k
    Garbage: 71002k -> 70958k
    Leak: 5995k -> 5994k
    Overhead: 8078k -> 8068k
    GGC runs: 507 -> 508
    Pre-IPA-Garbage: 14079k -> 14091k
    Pre-IPA-Leak: 17155k -> 17160k
    Pre-IPA-Overhead: 2133k -> 2134k
    Post-IPA-Garbage: 14079k -> 14091k
    Post-IPA-Leak: 17155k -> 17160k
    Post-IPA-Overhead: 2133k -> 2134k

comparing combine.c compilation at -O3 level:
    Overall memory needed: 38468k -> 38469k
    Peak memory use before GGC: 16500k -> 16506k
    Peak memory use after GGC: 16202k -> 16207k
    Maximum of released memory in single GGC run: 2010k -> 2012k
    Garbage: 87962k -> 87960k
    Leak: 6047k -> 6038k
    Overhead: 10015k -> 10009k
    GGC runs: 544 -> 545
    Pre-IPA-Garbage: 14081k -> 14093k
    Pre-IPA-Leak: 17167k -> 17172k
    Pre-IPA-Overhead: 2134k -> 2136k
    Post-IPA-Garbage: 14081k -> 14093k
    Post-IPA-Leak: 17167k -> 17172k
    Post-IPA-Overhead: 2134k -> 2136k

comparing insn-attrtab.c compilation at -O0 level:
    Overall memory needed: 137428k -> 137437k
    Peak memory use before GGC: 57005k
    Peak memory use after GGC: 31462k
    Maximum of released memory in single GGC run: 33256k
    Garbage: 128634k
    Leak: 8368k
    Overhead: 14369k
    GGC runs: 292
Testing has produced no results
Testing has produced no results

comparing insn-attrtab.c compilation at -O0 -g level:
    Overall memory needed: 138684k -> 138677k
    Peak memory use before GGC: 58147k
    Peak memory use after GGC: 32604k
    Maximum of released memory in single GGC run: 33256k
    Garbage: 128886k
    Leak: 10043k
    Overhead: 14724k
    GGC runs: 288
Testing has produced no results
Testing has produced no results

comparing insn-attrtab.c compilation at -O1 level:
    Overall memory needed: 142840k -> 142861k
    Peak memory use before GGC: 56207k
    Peak memory use after GGC: 49970k
    Maximum of released memory in single GGC run: 23976k
    Garbage: 205071k -> 205213k
    Leak: 9261k
    Overhead: 24747k -> 24762k
    GGC runs: 315
    Pre-IPA-Garbage: 50376k
    Pre-IPA-Leak: 49662k
    Pre-IPA-Overhead: 7543k
    Post-IPA-Garbage: 50376k
    Post-IPA-Leak: 49662k
    Post-IPA-Overhead: 7543k

comparing insn-attrtab.c compilation at -O2 level:
  Amount of produced GGC garbage increased from 245611k to 245878k, overall 0.11%
    Overall memory needed: 184388k -> 184565k
    Peak memory use before GGC: 56838k
    Peak memory use after GGC: 51558k
    Maximum of released memory in single GGC run: 22891k
    Garbage: 245611k -> 245878k
    Leak: 10345k
    Overhead: 30385k -> 30414k
    GGC runs: 346
    Pre-IPA-Garbage: 50448k
    Pre-IPA-Leak: 49666k
    Pre-IPA-Overhead: 7551k
    Post-IPA-Garbage: 50448k
    Post-IPA-Leak: 49666k
    Post-IPA-Overhead: 7551k

comparing insn-attrtab.c compilation at -O3 level:
    Overall memory needed: 190236k -> 190413k
    Peak memory use before GGC: 66384k
    Peak memory use after GGC: 62267k
    Maximum of released memory in single GGC run: 23391k
    Garbage: 269315k -> 269582k
    Leak: 10383k
    Overhead: 32173k -> 32203k
    GGC runs: 346
    Pre-IPA-Garbage: 50448k
    Pre-IPA-Leak: 49666k
    Pre-IPA-Overhead: 7551k
    Post-IPA-Garbage: 50448k
    Post-IPA-Leak: 49666k
    Post-IPA-Overhead: 7551k

comparing Gerald's testcase PR8361 compilation at -O0 level:
    Overall memory needed: 148705k -> 148743k
    Peak memory use before GGC: 84449k
    Peak memory use after GGC: 83613k
    Maximum of released memory in single GGC run: 16993k
    Garbage: 200804k
    Leak: 48036k
    Overhead: 23982k
    GGC runs: 417
    Pre-IPA-Garbage: 109751k
    Pre-IPA-Leak: 71096k
    Pre-IPA-Overhead: 12096k
    Post-IPA-Garbage: 109751k
    Post-IPA-Leak: 71096k
    Post-IPA-Overhead: 12096k

comparing Gerald's testcase PR8361 compilation at -O0 -g level:
    Overall memory needed: 169317k -> 169315k
    Peak memory use before GGC: 96105k
    Peak memory use after GGC: 95153k
    Maximum of released memory in single GGC run: 17492k
    Garbage: 206447k
    Leak: 69718k
    Overhead: 28959k
    GGC runs: 390
    Pre-IPA-Garbage: 110395k
    Pre-IPA-Leak: 84377k
    Pre-IPA-Overhead: 14672k
    Post-IPA-Garbage: 110395k
    Post-IPA-Leak: 84377k
    Post-IPA-Overhead: 14672k

comparing Gerald's testcase PR8361 compilation at -O1 level:
    Overall memory needed: 122239k -> 122241k
    Peak memory use before GGC: 83944k
    Peak memory use after GGC: 83112k
    Maximum of released memory in single GGC run: 16213k
    Garbage: 288569k -> 288582k
    Leak: 47318k -> 47318k
    Overhead: 29235k -> 29237k
    GGC runs: 512
    Pre-IPA-Garbage: 154817k -> 154817k
    Pre-IPA-Leak: 87733k
    Pre-IPA-Overhead: 17481k
    Post-IPA-Garbage: 154817k -> 154817k
    Post-IPA-Leak: 87733k
    Post-IPA-Overhead: 17481k

comparing Gerald's testcase PR8361 compilation at -O2 level:
    Overall memory needed: 124711k -> 124709k
    Peak memory use before GGC: 84160k
    Peak memory use after GGC: 83312k
    Maximum of released memory in single GGC run: 16215k
    Garbage: 354389k -> 354415k
    Leak: 48411k
    Overhead: 36234k -> 36237k
    GGC runs: 585
    Pre-IPA-Garbage: 157364k -> 157364k
    Pre-IPA-Leak: 87884k
    Pre-IPA-Overhead: 17762k
    Post-IPA-Garbage: 157364k -> 157364k
    Post-IPA-Leak: 87884k
    Post-IPA-Overhead: 17762k

comparing Gerald's testcase PR8361 compilation at -O3 level:
    Overall memory needed: 127599k -> 127609k
    Peak memory use before GGC: 85427k -> 85428k
    Peak memory use after GGC: 84551k
    Maximum of released memory in single GGC run: 16587k
    Garbage: 378189k -> 378225k
    Leak: 48997k
    Overhead: 38388k -> 38390k
    GGC runs: 610
    Pre-IPA-Garbage: 160081k -> 160079k
    Pre-IPA-Leak: 88167k
    Pre-IPA-Overhead: 18011k -> 18011k
    Post-IPA-Garbage: 160081k -> 160079k
    Post-IPA-Leak: 88167k
    Post-IPA-Overhead: 18011k -> 18011k

comparing PR rtl-optimization/28071 testcase compilation at -O0 level:
    Overall memory needed: 380450k -> 380452k
    Peak memory use before GGC: 101215k
    Peak memory use after GGC: 56868k
    Maximum of released memory in single GGC run: 50583k
    Garbage: 178905k
    Leak: 6103k
    Overhead: 30783k
    GGC runs: 107
Testing has produced no results
Testing has produced no results

comparing PR rtl-optimization/28071 testcase compilation at -O0 -g level:
    Overall memory needed: 381242k -> 381240k
    Peak memory use before GGC: 101845k
    Peak memory use after GGC: 57499k
    Maximum of released memory in single GGC run: 50582k
    Garbage: 179020k
    Leak: 7833k
    Overhead: 31215k
    GGC runs: 111
Testing has produced no results
Testing has produced no results

comparing PR rtl-optimization/28071 testcase compilation at -O1 level:
    Overall memory needed: 437188k -> 437281k
    Peak memory use before GGC: 78692k
    Peak memory use after GGC: 70441k
    Maximum of released memory in single GGC run: 37596k
    Garbage: 234583k
    Leak: 15907k
    Overhead: 35646k
    GGC runs: 106
    Pre-IPA-Garbage: 52926k
    Pre-IPA-Leak: 49238k
    Pre-IPA-Overhead: 7120k
    Post-IPA-Garbage: 52926k
    Post-IPA-Leak: 49238k
    Post-IPA-Overhead: 7120k

comparing PR rtl-optimization/28071 testcase compilation at -O2 level:
    Overall memory needed: 358144k -> 357945k
    Peak memory use before GGC: 75428k
    Peak memory use after GGC: 70442k
    Maximum of released memory in single GGC run: 32234k
    Garbage: 245354k -> 245359k
    Leak: 16079k
    Overhead: 38822k -> 38822k
    GGC runs: 117
    Pre-IPA-Garbage: 91838k
    Pre-IPA-Leak: 77295k
    Pre-IPA-Overhead: 11793k
    Post-IPA-Garbage: 91838k
    Post-IPA-Leak: 77295k
    Post-IPA-Overhead: 11793k

comparing PR rtl-optimization/28071 testcase compilation at -O3 -fno-tree-pre -fno-tree-fre level:
    Overall memory needed: 1197600k -> 1197633k
    Peak memory use before GGC: 136589k
    Peak memory use after GGC: 127947k
    Maximum of released memory in single GGC run: 59910k
    Garbage: 369974k
    Leak: 24893k
    Overhead: 49889k
    GGC runs: 104
    Pre-IPA-Garbage: 91838k
    Pre-IPA-Leak: 77295k
    Pre-IPA-Overhead: 11793k
    Post-IPA-Garbage: 91838k
    Post-IPA-Leak: 77295k
    Post-IPA-Overhead: 11793k

Head of the ChangeLog is:

--- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog	2008-03-18 15:19:35.000000000 +0000
+++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/ChangeLog	2008-03-18 18:58:12.000000000 +0000
@@ -1,3 +1,52 @@
+2008-03-18  Paolo Bonzini  <bonzini@gnu.org>
+
+        * expr.c (store_expr): Assume lang_hooks.reduce_bit_field_operations
+        is true.
+        (expand_expr_real_1) <REDUCE_BIT_FIELD>: Don't look at ignore.
+        (expand_expr_real_1): Assume lang_hooks.reduce_bit_field_operations
+        is true.  Add "&& !ignore" condition to reduce_bit_field.  Modify
+        target after ignore has been set, and move there also the commputation
+        of subtarget and original_target.
+        * langhooks-def.h (LANG_HOOKS_REDUCE_BIT_FIELD_OPERATIONS): Delete.
+        (LANG_HOOKS_INITIALIZER): Remove it.
+        * langhooks.h (struct lang_hooks): Remove reduce_bit_field_operations.
+
+2008-03-18  Richard Guenther  <rguenther@suse.de>
+
+	* tree-ssa-sccvn.c (visit_reference_op_load): If the lookup
+	found an expression with constants, note that in the VN for the lhs.
+	* tree-ssa-pre.c (eliminate): Visit COND_EXPR statements and
+	fold them to constants if possible.  Run cleanup_cfg if done so.
+	(execute_pre): Return todo.
+	(do_pre): Likewise.
+	(execute_fre): Likewise.
+	* tree-ssa-forwprop.c (can_propagate_from): Allow propagation
+	of constants.
+	(get_prop_source_stmt): Look through pointer conversions.
+
+2008-03-18  Jan Hubicka  <jh@suse.cz>
+
+	* tree-pretty-print.c: Include predict.h.
+	(dump_generic_node): Dump predictor.
+	* tree.h (PREDICT_EXPR_OUTCOME, PREDICT_EXPR_PREDICTION): Update.
+	* tree-gimple.c (is_gimple_stmt): Add PREDICT_EXPR.
+	* gimple-low.c (lower_stmt): Likewise.
+	* expr.c (expand_expr_real): Likewise.
+	* predict.c (tree_bb_level_predictions): Use PREDICT_EXPRs and remove
+	them.
+	(build_predict_expr, build_predict_expr): New.
+	* predict.h (predictor_name, build_predict_expr): Update.
+	* c-typeck.c (c_finish_bc_stmt): Add prediction.
+	* gimplify.c (gimplify_expr): Add PREDICT_EXPR.
+	* predict.def (PRED_CONTINUE): Update hitrate.
+	* tree.def (PREDICT_EXPR): Define.
+	* tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark PREDICT_EXPR;
+	do not handle BIND_EXPR.
+	* tree-inline.c (estimate_num_insns_1): PREDICT_EXPR is free.
+	* tree-cfg.c (verify_gimple_stmt): PREDICT_EXPR is valid.
+	* tree-ssa-operands.c (get_expr_operands): PREDICT_EXPR takes no
+	operands.
+
 2008-03-18  Michael Matz  <matz@suse.de>
 
 	* gcov-io.h (__gcov_merge_ior, __gcov_fork): Mark hidden.
@@ -23,6 +72,29 @@
 
 2008-03-17  Richard Guenther  <rguenther@suse.de>
 
+	* tree-pretty-print.c: Include predict.h.
+	(dump_generic_node): Dump predictor.
+	* tree.h (PREDICT_EXPR_OUTCOME, PREDICT_EXPR_PREDICTION): Update.
+	* tree-gimple.c (is_gimple_stmt): Add PREDICT_EXPR.
+	* gimple-low.c (lower_stmt): Likewise.
+	* expr.c (expand_expr_real): Likewise.
+	* predict.c (tree_bb_level_predictions): Use PREDICT_EXPRs and remove
+	them.
+	(build_predict_expr, build_predict_expr): New.
+	* predict.h (predictor_name, build_predict_expr): Update.
+	* c-typeck.c (c_finish_bc_stmt): Add prediction.
+	* gimplify.c (gimplify_expr): Add PREDICT_EXPR.
+	* predict.def (PRED_CONTINUE): Update hitrate.
+	* tree.def (PREDICT_EXPR): Define.
+	* tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark PREDICT_EXPR;
+	do not handle BIND_EXPR.
+	* tree-inline.c (estimate_num_insns_1): PREDICT_EXPR is free.
+	* tree-cfg.c (verify_gimple_stmt): PREDICT_EXPR is valid.
+	* tree-ssa-operands.c (get_expr_operands): PREDICT_EXPR takes no
+	operands.
+
+2008-03-17  Richard Guenther  <rguenther@suse.de>
+
 	PR tree-optimization/19637
 	* fold-const.c (fold_unary): Remove restrictions of removing
 	intermediate pointer-conversions (P2)(P1)P0.
--- /usr/src/SpecTests/sandbox-britten-memory/x86_64/mem-result/ChangeLog.cp	2008-03-18 04:13:38.000000000 +0000
+++ /usr/src/SpecTests/sandbox-britten-memory/gcc/gcc/cp/ChangeLog	2008-03-18 18:58:11.000000000 +0000
@@ -1,3 +1,7 @@
+2008-03-18  Paolo Bonzini  <bonzini@gnu.org>
+
+        * cp-lang.c (LANG_HOOKS_REDUCE_BIT_FIELD_OPERATIONS): Delete.
+	
 2008-03-17  Jason Merrill  <jason@redhat.com>
 
 	PR c++/35548


The results can be reproduced by building a compiler with

--enable-gather-detailed-mem-stats targetting x86-64

and compiling preprocessed combine.c or testcase from PR8632 with:

-fmem-report --param=ggc-min-heapsize=1024 --param=ggc-min-expand=1 -Ox -Q

The memory consumption summary appears in the dump after detailed listing
of the places they are allocated in.  Peak memory consumption is actually
computed by looking for maximal value in {GC XXXX -> YYYY} report.

Your testing script.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]